[FFmpeg-cvslog] hevc: Separate adding residual to prediction from IDCT

Alexandra Hájková git at videolan.org
Tue Jan 31 16:39:45 EET 2017


ffmpeg | branch: master | Alexandra Hájková <alexandra.khirnova at gmail.com> | Sun Jul  3 17:38:56 2016 +0200| [1bd890ad173d79e7906c5e1d06bf0a06cca4519d] | committer: Luca Barbato

hevc: Separate adding residual to prediction from IDCT

Based on patch 250430bf28118cf843df887e8c8b345f1c60c82d
by Mickaël Raulet <mraulet at insa-rennes.fr>, integrated
to Libav by Josh de Kock <josh at itanimul.li>.

Signed-off-by: Alexandra Hájková <alexandra at khirnov.net>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1bd890ad173d79e7906c5e1d06bf0a06cca4519d
---

 libavcodec/hevc.c             | 11 +++---
 libavcodec/hevcdsp.c          | 20 +++++-----
 libavcodec/hevcdsp.h          | 10 ++---
 libavcodec/hevcdsp_template.c | 89 ++++++++++++++-----------------------------
 4 files changed, 47 insertions(+), 83 deletions(-)

diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index 5e1b3a0..d5d3f59 100644
--- a/libavcodec/hevc.c
+++ b/libavcodec/hevc.c
@@ -1208,17 +1208,16 @@ static void hls_residual_coding(HEVCContext *s, int x0, int y0,
         }
     }
 
-    if (lc->cu.cu_transquant_bypass_flag) {
-        s->hevcdsp.transquant_bypass[log2_trafo_size - 2](dst, coeffs, stride);
-    } else {
+    if (!lc->cu.cu_transquant_bypass_flag) {
         if (transform_skip_flag)
-            s->hevcdsp.transform_skip(dst, coeffs, stride);
+            s->hevcdsp.dequant(coeffs);
         else if (lc->cu.pred_mode == MODE_INTRA && c_idx == 0 &&
                  log2_trafo_size == 2)
-            s->hevcdsp.transform_4x4_luma_add(dst, coeffs, stride);
+            s->hevcdsp.transform_4x4_luma(coeffs);
         else
-            s->hevcdsp.transform_add[log2_trafo_size - 2](dst, coeffs, stride);
+            s->hevcdsp.idct[log2_trafo_size - 2](coeffs);
     }
+    s->hevcdsp.add_residual[log2_trafo_size - 2](dst, coeffs, stride);
 }
 
 static int hls_transform_unit(HEVCContext *s, int x0, int y0,
diff --git a/libavcodec/hevcdsp.c b/libavcodec/hevcdsp.c
index 15a712d..4bd3d97 100644
--- a/libavcodec/hevcdsp.c
+++ b/libavcodec/hevcdsp.c
@@ -164,16 +164,16 @@ void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth)
 
 #define HEVC_DSP(depth)                                                     \
     hevcdsp->put_pcm                = FUNC(put_pcm, depth);                 \
-    hevcdsp->transquant_bypass[0]   = FUNC(transquant_bypass4x4, depth);    \
-    hevcdsp->transquant_bypass[1]   = FUNC(transquant_bypass8x8, depth);    \
-    hevcdsp->transquant_bypass[2]   = FUNC(transquant_bypass16x16, depth);  \
-    hevcdsp->transquant_bypass[3]   = FUNC(transquant_bypass32x32, depth);  \
-    hevcdsp->transform_skip         = FUNC(transform_skip, depth);          \
-    hevcdsp->transform_4x4_luma_add = FUNC(transform_4x4_luma_add, depth);  \
-    hevcdsp->transform_add[0]       = FUNC(transform_4x4_add, depth);       \
-    hevcdsp->transform_add[1]       = FUNC(transform_8x8_add, depth);       \
-    hevcdsp->transform_add[2]       = FUNC(transform_16x16_add, depth);     \
-    hevcdsp->transform_add[3]       = FUNC(transform_32x32_add, depth);     \
+    hevcdsp->add_residual[0]        = FUNC(add_residual4x4, depth);         \
+    hevcdsp->add_residual[1]        = FUNC(add_residual8x8, depth);         \
+    hevcdsp->add_residual[2]        = FUNC(add_residual16x16, depth);       \
+    hevcdsp->add_residual[3]        = FUNC(add_residual32x32, depth);       \
+    hevcdsp->dequant                = FUNC(dequant, depth);                 \
+    hevcdsp->transform_4x4_luma     = FUNC(transform_4x4_luma, depth);      \
+    hevcdsp->idct[0]                = FUNC(idct_4x4, depth);                \
+    hevcdsp->idct[1]                = FUNC(idct_8x8, depth);                \
+    hevcdsp->idct[2]                = FUNC(idct_16x16, depth);              \
+    hevcdsp->idct[3]                = FUNC(idct_32x32, depth);              \
                                                                             \
     hevcdsp->sao_band_filter[0] = FUNC(sao_band_filter_0, depth);           \
     hevcdsp->sao_band_filter[1] = FUNC(sao_band_filter_1, depth);           \
diff --git a/libavcodec/hevcdsp.h b/libavcodec/hevcdsp.h
index 4097233..decd1c9 100644
--- a/libavcodec/hevcdsp.h
+++ b/libavcodec/hevcdsp.h
@@ -42,13 +42,11 @@ typedef struct HEVCDSPContext {
     void (*put_pcm)(uint8_t *dst, ptrdiff_t stride, int size,
                     GetBitContext *gb, int pcm_bit_depth);
 
-    void (*transquant_bypass[4])(uint8_t *dst, int16_t *coeffs,
-                                 ptrdiff_t stride);
+    void (*add_residual[4])(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
 
-    void (*transform_skip)(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-    void (*transform_4x4_luma_add)(uint8_t *dst, int16_t *coeffs,
-                                   ptrdiff_t stride);
-    void (*transform_add[4])(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
+    void (*dequant)(int16_t *coeffs);
+    void (*transform_4x4_luma)(int16_t *coeffs);
+    void (*idct[4])(int16_t *coeffs);
 
     void (*sao_band_filter[4])(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
                                struct SAOParams *sao, int *borders,
diff --git a/libavcodec/hevcdsp_template.c b/libavcodec/hevcdsp_template.c
index 31a2e7a..b4816db 100644
--- a/libavcodec/hevcdsp_template.c
+++ b/libavcodec/hevcdsp_template.c
@@ -40,8 +40,8 @@ static void FUNC(put_pcm)(uint8_t *_dst, ptrdiff_t stride, int size,
     }
 }
 
-static av_always_inline void FUNC(transquant_bypass)(uint8_t *_dst, int16_t *coeffs,
-                                                     ptrdiff_t stride, int size)
+static av_always_inline void FUNC(add_residual)(uint8_t *_dst, int16_t *res,
+                                                ptrdiff_t stride, int size)
 {
     int x, y;
     pixel *dst = (pixel *)_dst;
@@ -50,41 +50,39 @@ static av_always_inline void FUNC(transquant_bypass)(uint8_t *_dst, int16_t *coe
 
     for (y = 0; y < size; y++) {
         for (x = 0; x < size; x++) {
-            dst[x] = av_clip_pixel(dst[x] + *coeffs);
-            coeffs++;
+            dst[x] = av_clip_pixel(dst[x] + *res);
+            res++;
         }
         dst += stride;
     }
 }
 
-static void FUNC(transquant_bypass4x4)(uint8_t *_dst, int16_t *coeffs,
-                                       ptrdiff_t stride)
+static void FUNC(add_residual4x4)(uint8_t *_dst, int16_t *res,
+                                  ptrdiff_t stride)
 {
-    FUNC(transquant_bypass)(_dst, coeffs, stride, 4);
+    FUNC(add_residual)(_dst, res, stride, 4);
 }
 
-static void FUNC(transquant_bypass8x8)(uint8_t *_dst, int16_t *coeffs,
-                                       ptrdiff_t stride)
+static void FUNC(add_residual8x8)(uint8_t *_dst, int16_t *res,
+                                  ptrdiff_t stride)
 {
-    FUNC(transquant_bypass)(_dst, coeffs, stride, 8);
+    FUNC(add_residual)(_dst, res, stride, 8);
 }
 
-static void FUNC(transquant_bypass16x16)(uint8_t *_dst, int16_t *coeffs,
-                                         ptrdiff_t stride)
+static void FUNC(add_residual16x16)(uint8_t *_dst, int16_t *res,
+                                    ptrdiff_t stride)
 {
-    FUNC(transquant_bypass)(_dst, coeffs, stride, 16);
+    FUNC(add_residual)(_dst, res, stride, 16);
 }
 
-static void FUNC(transquant_bypass32x32)(uint8_t *_dst, int16_t *coeffs,
-                                         ptrdiff_t stride)
+static void FUNC(add_residual32x32)(uint8_t *_dst, int16_t *res,
+                                    ptrdiff_t stride)
 {
-    FUNC(transquant_bypass)(_dst, coeffs, stride, 32);
+    FUNC(add_residual)(_dst, res, stride, 32);
 }
 
-static void FUNC(transform_skip)(uint8_t *_dst, int16_t *coeffs,
-                                 ptrdiff_t stride)
+static void FUNC(dequant)(int16_t *coeffs)
 {
-    pixel *dst = (pixel *)_dst;
     int shift  = 13 - BIT_DEPTH;
 #if BIT_DEPTH <= 13
     int offset = 1 << (shift - 1);
@@ -93,19 +91,14 @@ static void FUNC(transform_skip)(uint8_t *_dst, int16_t *coeffs,
 #endif
     int x, y;
 
-    stride /= sizeof(pixel);
-
     for (y = 0; y < 4 * 4; y += 4) {
         for (x = 0; x < 4; x++)
-            dst[x] = av_clip_pixel(dst[x] + ((coeffs[y + x] + offset) >> shift));
-        dst += stride;
+            coeffs[y + x] = (coeffs[y + x] + offset) >> shift;
     }
 }
 
 #define SET(dst, x)   (dst) = (x)
 #define SCALE(dst, x) (dst) = av_clip_int16(((x) + add) >> shift)
-#define ADD_AND_SCALE(dst, x)                                           \
-    (dst) = av_clip_pixel((dst) + av_clip_int16(((x) + add) >> shift))
 
 #define TR_4x4_LUMA(dst, src, step, assign)                             \
     do {                                                                \
@@ -122,17 +115,13 @@ static void FUNC(transform_skip)(uint8_t *_dst, int16_t *coeffs,
         assign(dst[3 * step], 55 * c0 + 29 * c2 - c3);                  \
     } while (0)
 
-static void FUNC(transform_4x4_luma_add)(uint8_t *_dst, int16_t *coeffs,
-                                         ptrdiff_t stride)
+static void FUNC(transform_4x4_luma)(int16_t *coeffs)
 {
     int i;
-    pixel *dst   = (pixel *)_dst;
     int shift    = 7;
     int add      = 1 << (shift - 1);
     int16_t *src = coeffs;
 
-    stride /= sizeof(pixel);
-
     for (i = 0; i < 4; i++) {
         TR_4x4_LUMA(src, src, 4, SCALE);
         src++;
@@ -141,9 +130,8 @@ static void FUNC(transform_4x4_luma_add)(uint8_t *_dst, int16_t *coeffs,
     shift = 20 - BIT_DEPTH;
     add   = 1 << (shift - 1);
     for (i = 0; i < 4; i++) {
-        TR_4x4_LUMA(dst, coeffs, 1, ADD_AND_SCALE);
+        TR_4x4_LUMA(coeffs, coeffs, 1, SCALE);
         coeffs += 4;
-        dst    += stride;
     }
 }
 
@@ -166,17 +154,13 @@ static void FUNC(transform_4x4_luma_add)(uint8_t *_dst, int16_t *coeffs,
         assign(dst[3 * dstep], e0 - o0);                                \
     } while (0)
 
-static void FUNC(transform_4x4_add)(uint8_t *_dst, int16_t *coeffs,
-                                    ptrdiff_t stride)
+static void FUNC(idct_4x4)(int16_t *coeffs)
 {
     int i;
-    pixel *dst   = (pixel *)_dst;
     int shift    = 7;
     int add      = 1 << (shift - 1);
     int16_t *src = coeffs;
 
-    stride /= sizeof(pixel);
-
     for (i = 0; i < 4; i++) {
         TR_4(src, src, 4, 4, SCALE);
         src++;
@@ -185,9 +169,8 @@ static void FUNC(transform_4x4_add)(uint8_t *_dst, int16_t *coeffs,
     shift = 20 - BIT_DEPTH;
     add   = 1 << (shift - 1);
     for (i = 0; i < 4; i++) {
-        TR_4(dst, coeffs, 1, 1, ADD_AND_SCALE);
+        TR_4(coeffs, coeffs, 1, 1, SCALE);
         coeffs += 4;
-        dst    += stride;
     }
 }
 
@@ -241,17 +224,13 @@ static void FUNC(transform_4x4_add)(uint8_t *_dst, int16_t *coeffs,
 
 
 
-static void FUNC(transform_8x8_add)(uint8_t *_dst, int16_t *coeffs,
-                                    ptrdiff_t stride)
+static void FUNC(idct_8x8)(int16_t *coeffs)
 {
     int i;
-    pixel *dst   = (pixel *)_dst;
     int shift    = 7;
     int add      = 1 << (shift - 1);
     int16_t *src = coeffs;
 
-    stride /= sizeof(pixel);
-
     for (i = 0; i < 8; i++) {
         TR_8(src, src, 8, 8, SCALE);
         src++;
@@ -260,23 +239,18 @@ static void FUNC(transform_8x8_add)(uint8_t *_dst, int16_t *coeffs,
     shift = 20 - BIT_DEPTH;
     add   = 1 << (shift - 1);
     for (i = 0; i < 8; i++) {
-        TR_8(dst, coeffs, 1, 1, ADD_AND_SCALE);
+        TR_8(coeffs, coeffs, 1, 1, SCALE);
         coeffs += 8;
-        dst    += stride;
     }
 }
 
-static void FUNC(transform_16x16_add)(uint8_t *_dst, int16_t *coeffs,
-                                      ptrdiff_t stride)
+static void FUNC(idct_16x16)(int16_t *coeffs)
 {
     int i;
-    pixel *dst   = (pixel *)_dst;
     int shift    = 7;
     int add      = 1 << (shift - 1);
     int16_t *src = coeffs;
 
-    stride /= sizeof(pixel);
-
     for (i = 0; i < 16; i++) {
         TR_16(src, src, 16, 16, SCALE);
         src++;
@@ -285,23 +259,18 @@ static void FUNC(transform_16x16_add)(uint8_t *_dst, int16_t *coeffs,
     shift = 20 - BIT_DEPTH;
     add   = 1 << (shift - 1);
     for (i = 0; i < 16; i++) {
-        TR_16(dst, coeffs, 1, 1, ADD_AND_SCALE);
+        TR_16(coeffs, coeffs, 1, 1, SCALE);
         coeffs += 16;
-        dst    += stride;
     }
 }
 
-static void FUNC(transform_32x32_add)(uint8_t *_dst, int16_t *coeffs,
-                                      ptrdiff_t stride)
+static void FUNC(idct_32x32)(int16_t *coeffs)
 {
     int i;
-    pixel *dst   = (pixel *)_dst;
     int shift    = 7;
     int add      = 1 << (shift - 1);
     int16_t *src = coeffs;
 
-    stride /= sizeof(pixel);
-
     for (i = 0; i < 32; i++) {
         TR_32(src, src, 32, 32, SCALE);
         src++;
@@ -310,9 +279,8 @@ static void FUNC(transform_32x32_add)(uint8_t *_dst, int16_t *coeffs,
     shift = 20 - BIT_DEPTH;
     add   = 1 << (shift - 1);
     for (i = 0; i < 32; i++) {
-        TR_32(dst, coeffs, 1, 1, ADD_AND_SCALE);
+        TR_32(coeffs, coeffs, 1, 1, SCALE);
         coeffs += 32;
-        dst    += stride;
     }
 }
 
@@ -769,7 +737,6 @@ static void FUNC(sao_edge_filter_3)(uint8_t *_dst, uint8_t *_src,
 
 #undef SET
 #undef SCALE
-#undef ADD_AND_SCALE
 #undef TR_4
 #undef TR_8
 #undef TR_16



More information about the ffmpeg-cvslog mailing list